Beyond the Matrix-Assisted Laser Desorption Ionization (MALDI)Biotyping Workflow: in Search of Microorganism-Specific TrypticPeptides Enabling Discrimination of Subspecies

Maria-Theresia Gekenidis,a Patrick Studer,a* Simone Wüthrich,b René Brunisholz,b David Drissnera

Agroscope, Institute for Food Sciences, Waedenswil, Switzerlanda; Functional Genomics Center Zurich, ETH Zurich and University of Zurich, Zurich, Switzerlandb

A well-accepted method for identification of microorganisms uses matrix-assisted laser desorption ionization–time of flightmass spectrometry (MALDI-TOF MS) coupled to analysis software which identifies and classifies the organism according to itsribosomal protein spectral profile. The method, called MALDI biotyping, is widely used in clinical diagnostics and has partlyreplaced conventional microbiological techniques such as biochemical identification due to its shorter time to result (minutesfor MALDI biotyping versus hours or days for classical phenotypic or genotypic identification). Besides its utility for identifyingbacteria, MS-based identification has been shown to be applicable also to yeasts and molds. A limitation to this method, how-ever, is that accurate identification is most reliably achieved on the species level on the basis of reference mass spectra, makingfurther phylogenetic classification unreliable. Here, it is shown that combining tryptic digestion of the acid/organic solvent ex-tracted (classical biotyping preparation) and resolubilized proteins, nano-liquid chromatography (nano-LC), and subsequentidentification of the peptides by MALDI-tandem TOF (MALDI-TOF/TOF) mass spectrometry increases the discriminationpower to the level of subspecies. As a proof of concept, using this targeted proteomics workflow, we have identified subspecies-specific biomarker peptides for three Salmonella subspecies, resulting in an extension of the mass range and type of proteins in-vestigated compared to classical MALDI biotyping. This method therefore offers rapid and cost-effective identification and clas-sification of microorganisms at a deeper taxonomic level.

Matrix-assisted laser desorption ionization (MALDI) biotyp-ing is based on unique ribosomal protein profiles matched

to a reference database and identified accordingly (1–5). The re-sulting fingerprints of these constantly expressed and highly abun-dant proteins are highly reproducible and mostly independent ofthe culture medium, incubation temperature, and growth state(6–8). MALDI biotyping has gained much attention in recentyears due to its speed, simple handling, cost-effectiveness, andhigh-throughput capabilities (9). Conventional methods assess-ing phenotypic traits based on biochemical reactions, antigen-antibody reactions, Gram staining, colony morphology, or growthpattern are often time-consuming and expensive. Molecular biol-ogy techniques for assessment of genotypic traits such as PCR,sequencing, pulsed-field gel electrophoresis (PFGE), multilocussequence typing (MLST), restriction fragment length polymor-phism (RFLP), and microarrays need great expertise and are ex-pensive as well. However, MALDI biotyping has limitations, too.The identification rates using clinical isolates are reported to be79.9% to 93.6% at the species level (10–13) and 94.5% to 97.2% atthe genus level (10, 12, 14). These rates are mainly due to highprotein identity among bacterial subspecies and serovars. There-fore, there is a need for a more sensitive and accurate method. Thisis particularly important in food safety, where contamination byfood-borne bacteria such as salmonellae can lead to serious illnessor even death and can cause economic and reputational losses toagriculture and the food industry. In most cases, contamination isintroduced during production, processing, or storage. The rapidbacterial profiling through MALDI-TOF MS has also found appli-cation in clinical microbiology, where characterization of certainspecies proved more accurate with MALDI biotyping than with16S rRNA gene sequencing (15). Other fields of application arebiodefense and environmental microbiology (16), where rapid

detection and differentiation of bacteria from surface, air, soil, orwater samples below the species level are key to evaluating theircommensal, mutualistic, or pathogenic characteristics. It is worthmentioning that the extent and quality of entries as well as thecorrect taxonomy of the microorganisms deposited in the data-base are crucial. Here we extend this method to the level of pep-tides. In this report, we present methods designed to improveultrafast generation of peptides that could act as Salmonella sub-species biomarkers. We investigate the use of high-intensity fo-cused ultrasound (HIFU) (17) to aid solubilization of the pelletedproteins after acid/organic solvent extraction and study the courseof tryptic digestion under HIFU conditions to optimize peptideoutput and generation of subspecies-specific peptide biomarkers.

MATERIALS AND METHODSBacterial culture and protein extraction (classical biotyping). Proteinextraction was performed in accordance with a procedure previously de-scribed (18) and lately applied by many research groups (19–26). Forclassical biotyping, which uses undigested cell extracts, Salmonella en-terica subsp. arizonae (German Collection of Microorganisms and Cell

Received 4 March 2014 Accepted 28 April 2014

Published ahead of print 2 May 2014

Editor: D. W. Schaffner

Address correspondence to David Drissner, david.drissner@agroscope.admin.ch.

* Present address: Patrick Studer, ETH Zurich, Institute of Food, Nutrition andHealth, Zurich, Switzerland.

Supplemental material for this article may be found at http://dx.doi.org/10.1128/AEM.00740-14.

Copyright © 2014, American Society for Microbiology. All Rights Reserved.

doi:10.1128/AEM.00740-14

4234 aem.asm.org Applied and Environmental Microbiology p. 4234 – 4241 July 2014 Volume 80 Number 14

on June 8, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Cultures; DSMZ 9386), S. enterica subsp. enterica (isolate from a gastro-enteritis outbreak in Scandinavia [27]), and S. enterica subsp. houtenae(DSMZ 9221) were grown overnight on tryptic soy agar (Sigma-Aldrich,St. Louis, MO). An inoculation loop full of cells was then introduced into1.2 ml of 75% ethanol (Sigma-Aldrich) and suspended by mixing with avortex device. The sample was centrifuged for 2 min at 16,100 � g, thesupernatant was discarded, and the residual ethanol was evaporated for 5min at room temperature. In order to extract the proteins, 50 �l of 70%formic acid (Sigma-Aldrich) and 50 �l of 100% acetonitrile (BiosolveB.V., Valkenswaard, Netherlands) were added to the pellet and thor-oughly mixed using a vortex device. Another centrifugation at 16,100 � gfor 2 min resulted in a protein-containing supernatant which was used forspotting onto the MALDI target (MSP 96 polished steel BC target; BrukerDaltonics, Bremen, Germany) and the subsequent MALDI-TOF mea-surement.

MALDI-TOF measurement. Protein-containing supernatant (1 �l)was spotted onto the MALDI target, allowed to dry, and covered with 1 �lof saturated matrix solution containing HCCA (�-cyano-4-hydroxycin-namic acid) (Bruker Daltonics) at a concentration of 10 mg/ml in aceto-nitrile-water-trifluoroacetic acid (TFA) (50:47.5:2.5 [vol/vol/vol]) (Sig-ma-Aldrich). Direct smearing of bacterial cells was performed in parallel:cell material from the agar plate was transferred onto the MALDI targetusing a toothpick, and the smear was covered with saturated HCCA ma-trix. Sample spectra were collected with a microflex LT MALDI-TOF massspectrometer (Bruker Daltonics) using flexControl software (Version 3.4)(Bruker Daltonics) and applying the measurement parameters suggestedby the manufacturer for classical biotyping: laser frequency of 60 Hz in thepositive linear mode with acquisition ranging from 2 to 20 kDa. Finalspectra consisted of 240 shots per spot (40 shots per raster spot). The laserintensity was chosen so as to obtain spectra with maximal absolute peakintensities ranging from about 5 � 103 to 104 arbitrary units. The spectrawere evaluated using the accompanying MALDI BioTyper OC software(Version 3.1) (Bruker Daltonics).

Bacterial culture and protein extraction (biomarker search). Forgeneration of biomarker peptides, an overnight culture of S. entericasubsp. arizonae, S. enterica subsp. enterica, and S. enterica subsp. houtenaegrown in tryptic soy broth (Sigma-Aldrich) was diluted to an optical den-sity at 600 nm (OD600) of 1. A 1-ml volume thereof was spun at 5,000 � gfor 1 min, and the resulting pellet was washed with deionized water. Aftera second spin, the pellet was resuspended in 1.2 ml of 75% ethanol andmixed using a vortex device. The sample was then centrifuged for 2 min at16,100 � g, the supernatant was discarded, and the residual ethanol wasevaporated for 5 min at room temperature. In order to extract the pro-teins, 100 �l of 70% formic acid and 100 �l of 100% acetonitrile wereadded to the pellet and thoroughly mixed using a vortex device. Afteranother centrifugation at 16,100 � g for 2 min, the protein-containingsupernatant was transferred to a new Eppendorf tube and put into thefreezer at �20°C until used for further preparation.

Trypsin digestion. The supernatant (175 �l) of the samples describedin the above section was subjected to complete evaporation using aSpeedVac concentrator (Savant Instruments Inc., New York, NY) atroom temperature. The resulting pellet was resuspended in 3 �l of100% acetonitrile, 3 �l of RapiGest (Waters, Guyancourt, France), 18�l of 100 mM Tris-HCl (Sigma-Aldrich) (pH 8.2), and 3 �l of 1 MTris-HCl (Sigma-Aldrich) (pH 8.2). This sample was then subjected to5 min of HIFU treatment (model UTR200, Hielscher Ultrasonics, Tel-tow, Germany) in order to resolubilize the pellet. Settings were asfollowing: intensity, 90%; cycles, 0.8. A 3-�l volume of a solution of 0.1mg/ml trypsin (Roche Diagnostics, Mannheim, Germany)–10 mMHCl (Merck, Darmstadt, Germany) was added subsequently, and theproteins were digested during a 15-min HIFU treatment (intensity,90%; cycles, 0.8) in an ice water bath.

Nano-LC spotting. The resulting peptides were separated with aneasy-nLC II nano-liquid chromatography (nano-LC) system (Bruker Dal-tonics) which was coupled to a Proteineer fc II fraction collector (Bruker

Daltonics). Mobile phase A consisted of water containing 0.1% TFA (Sig-ma-Aldrich), and mobile phase B consisted of 90% acetonitrile (Merck)containing 0.1% TFA. Separation was performed on a C18 column (Ac-claim PepMap100; Dionex Benelux B.V., Amsterdam, Netherlands) (15cm by 75 �m, 3 �m pore size, 100 Å) with a linear gradient of 2% to 45%mobile phase B in 64 min and a flow rate of 300 nl/min. A C18 reversed-phase solid-phase extraction trap (BioSphere NS-MP-10; NanoSepara-tions, Nieuwkoop, Netherlands) (2 cm by 100 �m, 5 �m, 120 Å) was usedto prevent the separation column from being jammed. The separatedpeptides were mixed with the HCCA matrix directly in the Proteineer fc IIfraction collector and spotted onto an MTP AnchorChip 1536 TF instru-ment (Bruker Daltonics) with an interval of six spots per min. The HCCAmatrix used for spotting was prepared as follows: 748 �l of TA95 (aceto-nitrile-water-TFA, 95:4.9:0.1 [vol/vol/vol]), 36 �l of saturated HCCA–TA90 (acetonitrile-water-TFA, 90:9.9:0.1 [vol/vol/vol]), 8 �l of 10% TFA,and 8 �l of 100 mM NH4H2PO4 (Sigma-Aldrich) dissolved in water. Oneper eight spots was manually spotted with peptide calibration standard II(Bruker Daltonics) diluted 1:200 in HCCA matrix prepared with 748 �l ofTA85 (acetonitrile-water-TFA, 85:14.9:0.1 [vol/vol/vol]) instead of TA95.

MALDI-TOF/TOF measurement. MALDI measurements of trypsin-digested samples were performed on an ultrafleXtreme MALDI-tan-dem TOF (MALDI-TOF/TOF) MS equipped with a SmartBeam laser(Bruker Daltonics). WARP-LC (Version 1.3) was used to set up dataacquisition methods, define values for tandem MS (MS/MS) precursorselection (signal-to-noise [S/N] ratio threshold, 10), and merge com-pounds that were separated by less than six fractions (mass tolerance,�50 ppm). The measurement parameters were again programmed inflexControl (Version 3.4) as follows: a laser frequency of 1,000 Hz inthe positive reflectron mode was used, with acquisition ranging from700 to 4,000 Da. Final spectra consisted of 3,000 shots per spot (100shots per raster spot). The laser intensity and detector sensitivity wereadjusted such that the highest peak of the spectrum was in the range of104 to 105 arbitrary units.

MS/MS data acquired through WARP-LC were searched on MASCOT(Version 2.4.1) using ProteinScape (Version 3.0; Bruker Daltonics) withthe search restricted to Salmonella on the NCBI database (status as of June2013). The search was limited to tryptic peptides with variable methio-nine, histidine, and tryptophan oxidation and an MS/MS toleranceof �0.7 Da. The peptide tolerance was set to �50 ppm. The peptide chargewas defined as �1, and one miscleavage was allowed. A peptide decoydatabase was used as well as the MASCOT Percolator algorithm to im-prove the significance of the search results. A second search was con-ducted on the UniProtKB/Swiss-Prot database, the results of which areshown in Table S1 in the supplemental material.

Data analysis. Data evaluation and a biomarker search were per-formed using a Microsoft Excel 2010 macro developed in house. Briefly,the flexAnalysis (Version 3.4) peak lists or WARP-LC compound listsfrom classical biotyping or protein digests, respectively, were exportedand merged in one Excel sheet. In a first step, an S/N ratio and a ppm valuewere defined. The values chosen for evaluation of the MS data were an S/Nratio of 3 and a ppm value of �200, whereas for the MS/MS data the S/Nratio was set to 10 and the effect of various ppm values (�25, �50, and�100 ppm) was investigated. The macro then selected data sets with anS/N ratio above the defined value and created a pivot table with the cor-responding m/z values in the row fields and the name of each of thedifferent measured sets in the column fields. Each m/z value was thentaken as the center and m/z values within the surrounding ppm windowwere counted. The counts of the names of the different measured setswithin the defined ppm window were the data items. This pivot table givesan overview of the number of m/z signals within the ppm window per setof subspecies. In a second step, m/z values of biomarkers were selected andpasted into a new Excel sheet. An m/z value was determined to be a bio-marker for a subspecies when it was present in all sets of the correspondingsubspecies but in none of the other subspecies.

Tryptic Peptides for Discrimination of Subspecies

July 2014 Volume 80 Number 14 aem.asm.org 4235

on June 8, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

RESULTSMALDI-TOF (classical biotyping). MALDI-TOF MS mass spec-tra were obtained for three different subspecies of S. enterica (S.enterica subsp. arizonae, S. enterica subsp. enterica, and S. entericasubsp. houtenae) by either directly smearing bacterial cells fromthe agar plate onto the MALDI target or spotting cell extracts. Forboth direct smearing and extraction, the experiment was repeatedthree times, each time using the same overnight culture for bothapproaches. Direct smearing yielded identification scores (datanot shown) that were slightly lower overall than those seen withthe cell extracts. Therefore, only the results obtained from the cellextracts are presented.

By classical biotyping according to the manufacturer’s proto-col, all three S. enterica samples could be correctly identified on thespecies level (Fig. 1A to C). However, identification on the sub-species level could not be achieved consistently.

MALDI-TOF/TOF (biomarker search). MALDI-TOF/TOFMS experiments were performed in triplicate using trypsin-di-gested cell extracts of the same three Salmonella samples. The op-timized workflow established in the current work is depicted inFig. 2. For that figure, tryptic digestion of resolubilized proteinsand nano-LC separation of the resulting peptides were combinedwith identification of those peptides by MALDI-TOF/TOF MS.The analysis of the acquired data by searching the NCBI databasefor each sample yielded a number of proteins, ranging from 465 to627 proteins.

Further data analysis with the Excel macro developed in houseyielded a list of potential biomarker peptide masses for the threesubspecies. The following searches were conducted. The S/N ratiowas set to 10 according to the value defined for MS/MS precursorselection, and specific ppm values (�25, �50, and �100 ppm) inthe range of the mass tolerance defined for MS/MS spectra acqui-sition were tested. An m/z value was classified as being a biomarkercandidate when it was present in all three sets of the correspondingsubspecies but in none of the other subspecies. Table 1 shows thenumber of potential biomarker peptide masses found for eachsubspecies as well as the number of masses assigned to actual tryp-tic peptides of Salmonella proteins found in the MASCOT searchand belonging to the correct subspecies (�200 ppm; S/N ratio �3). The analysis exhibited 51, 48, and 26 potential biomarker pep-tide masses for S. enterica subsp. arizonae, S. enterica subsp. en-terica, and S. enterica subsp. houtenae, respectively. Of thosemasses, about 50% were assigned to actual Salmonella peptides(Table 1).

In a second approach, the acquired data were analyzed bysearching the UniProtKB/Swiss-Prot database. Details on the pro-cedure are given in the supplemental material. All biomarker pep-tides identified for the three subspecies which were assigned toSalmonella proteins in this second MASCOT search are listed inTable S1 in the supplemental material.

Subspecies biomarker sequence comparison. For each sub-species, a sequence comparison of the assigned peptides to theanalogous peptide sequences of the other two subspecies was con-ducted, revealing the amino acid exchanges responsible for thesubspecies specificity of each peptide. Table 2 displays some se-lected examples from the NCBI search for biomarker peptides, allof which are distinctive for each of the three investigated subspe-cies (i.e., all three peptide sequences differ from one another). Forexample, a biomarker peptide in the isoprenoid biosynthesis pro-

tein of S. enterica subsp. houtenae (range, 73 to 99) shows aminoacid exchanges from serine to alanine in S. enterica subsp. arizonaeand from glutamic acid to aspartic acid in S. enterica subsp. en-terica. Of note, all biomarker peptides listed in Table 2 were foundat �25 ppm, except for the flagellar synthesis protein FlgN, whichwas detected only when the ppm value was set to �50 or �100.The ferric uptake regulator was additionally found in the �50-ppm search, and another nine biomarker peptide masses werefound in the �50- and �100-ppm searches (marked with an italicsuperscript “b” in Table 2). A complete overview of all potentialbiomarker peptides which were annotated to Salmonella proteinsin the NCBI database search, including the respective ppm values,can be found in the supplemental material (see Table S2 in thesupplemental material).

DISCUSSION

Identification of Salmonella on the genus level by classical MALDIbiotyping was reported earlier by Sparbier and colleagues (26). Inthe present study, the identification was correct on the specieslevel; however, the results were inconsistent on the subspecieslevel, the spectra being too similar to consistently distinguish thethree S. enterica subspecies (Fig. 1). The impression from the vi-sual spectra comparison was confirmed by a biomarker searchconducted with an Excel macro, analyzing peak lists generatedfrom spectra with the flexAnalysis software. With an S/N ratio of 3and a ppm value of �200, no characteristic biomarker peaks couldbe found for any of the three subspecies. The S/N ratio of 3 waschosen since the same value is used by the MALDI BioTyper OCsoftware for spectra evaluation. The selected ppm value comprisesthe approximate resolution range of the MALDI-TOF device ac-cording to its general specifications.

The approach to develop a workflow beyond the classicalMALDI biotyping (Fig. 2) in order to discriminate on the subspe-cies level revealed that about 10 times more proteins can be de-tected than with classical MALDI-TOF MS. The number of pro-teins found by classical MALDI biotyping ranged from 51 to 59 ina mass range of 3 to 14 kDa, including many ribosomal proteins(7). Applying the HIFU-assisted proteomics workflow clearly ex-tended the mass range and type of possible biomarker target pep-tides, e.g., the S. enterica subsp. houtenae-biomarker peptide de-tected in the 170-kDa cell division MukB protein (see Table S2 inthe supplemental material), and thus gave rise to more taxonomicdiscrimination power. The potential of a broader mass range toincrease discrimination power was shown for protein biomarkersof Gram-positive as well as Gram-negative bacteria from 15 to 75kDa by Madonna and coworkers (28).

The MALDI-TOF/TOF biomarker peptide search was con-ducted using an S/N ratio of 10 and ppm values of �25, �50,and �100. Most biomarker peptides were detected with morethan one ppm value. Of special interest are the masses which werefound with all three ppm values (entries marked with an italicsuperscript “b” in Table 2). Having been found at �25 ppm, themasses of the three replicates lie relatively close to one another,and the fact that they still appear as a biomarker in the �100-ppmsearch means that there are no relevant interfering peaks within arelatively broad mass window around them. They can therefore beconsidered more robust candidates than other biomarker pep-tides. More examples can be found in Tables S1 and S2 in thesupplemental material.

For some proteins which have been identified via the MALDI-

Gekenidis et al.

4236 aem.asm.org Applied and Environmental Microbiology

on June 8, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

FIG 1 MALDI-TOF MS mass spectra of cell extracts of S. enterica subsp. arizonae (A), S. enterica subsp. enterica (B), and S. enterica subsp. houtenae (C). Eachexperiment was performed in triplicate. The absolute intensities of the ions are shown on the y axis. The mass-to-charge (m/z) ratios of the ions are shown on thex axis, which correspond to the masses of the single positively charged ions.

July 2014 Volume 80 Number 14 aem.asm.org 4237

on June 8, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

TOF/TOF workflow (Fig. 2), more than one tryptic biomarkerpeptide could be found. These proteins are the DamX cell divisionprotein, 60-kDa chaperonin, NIpB/DapX lipoprotein, RNase E,and outer membrane protein A (see Table S2 in the supplementalmaterial). Also, a biomarker peptide was measured in some casesand annotated in more than one subspecies, making it an espe-cially robust candidate for subspecies distinction (marked with anasterisk in Table 2). Thus, the biomarker peptide in the isoprenoidbiosynthesis protein, which is distinctive for all three subspecies,was measured and annotated for each of the three, with averagemasses of 2,747.45 � 0.02, 2,749.45 � 0.01, and 2,763.46 � 0.02for S. enterica subsp. arizonae, S. enterica subsp. enterica, and S.enterica subsp. houtenae, respectively. Other examples are the fla-gellar synthesis protein FlgN, RNase E, N-ethylmaleimide reduc-tase, and outer membrane protein A (Table 2).

It is noteworthy that other working groups reported the iden-tification of Salmonella subspecies- or serovar-specific proteinbiomarkers as well. Dieckmann and coworkers (29) analyzed 126Salmonella strains by whole-cell MALDI-TOF MS for variations inprotein mass spectral profiles revealing protein biomarker peaksuseful for discrimination of S. enterica subspecies. We could con-firm two of the detected protein biomarker peaks, those corre-sponding to ribosomal protein L17 and glutaredoxin-1, to be spe-cific for S. enterica subsp. enterica. The responsible amino acidexchanges are listed in Table S1 of the supplemental material.

In a later study, Dieckmann and Malorny (30) screened 913 S.enterica subsp. enterica strains by whole-cell MALDI-TOF MS inorder to find serovar-specific biomarker proteins. They found theL17 ribosomal protein to be a major subspecies-specific bio-marker for S. enterica subsp. enterica, being present in almost all 89S. enterica subsp. enterica serovars tested and absent in all other S.enterica subspecies as well as S. bongori. The biomarker peptide

responsible for this subspecies specificity could be identified in thepresent study in the L17 50S ribosomal protein (see Table S1 in thesupplemental material), containing an additional serine residueand a threonine-to-alanine mutation for S. enterica subsp. arizo-nae and S. enterica subsp. houtenae, as stated by Dieckmann andMalorny and confirmed in the present work.

It is noteworthy that the abundance of existing serovars withina subspecies makes it difficult to ascertain that a peptide sequenceis identical in all serovars, especially when considering S. entericasubsp. enterica, for which above 2,500 serovars are known. Thus,Dieckmann and colleagues (29) found DNA-binding proteinH-NS (theoretical mass, 15,412.5 Da) to be a genus-identifyingbiomarker since it was present in all subspecies of S. enterica andeven in S. bongori. In a later study (30), however, the same proteinwas found to have a glycine-to-aspartic acid mutation yielding apredicted mass of 15,470.5 Da in S. enterica subsp. enterica serovarParatyphi A. In our studies, the H-NS histone family was found tocontain a biomarker peptide specific for S. enterica subspecies,with an exchange of methionine-aspartic acid for isoleucine-as-paragine in S. enterica subsp. arizonae and an aspartic acid-to-glycine mutation in S. enterica subsp. houtenae.

A very extensive review has been published recently on theapplication of MALDI-TOF MS profiling of bacteria at the strainlevel (16). Studies conducted so far cover a variety of microorgan-isms and have used different procedures for data evaluation,namely, library-based approaches and bioinformatics-enabledapproaches. Of special interest to the present study are the bioin-formatics-enabled approaches. For example, Fagerquist et al. (31)employed bottom-up proteomics to identify five species of Cam-pylobacter from protein extracts previously separated by high-per-formance LC (HPLC) and one-dimensional (1D) sodium dodecylsulfate polyacrylamide gel electrophoresis. They could identifyone protein biomarker, a DNA-binding protein, whose aminoacid sequence displayed intra- and interspecies variations, therebyfacilitating strain differentiation. In a follow-up study, the sameresearchers investigated three strains of Campylobacter jejuni (32)and found several protein biomarkers displaying mass shifts inMALDI-TOF MS profiles due to amino acid exchanges ratherthan posttranslational modifications. As in the present work,MASCOT analysis was used to confirm protein identities.

In 2010, the same group could distinguish two pathogenicO157:H7 strains from one non-O157:H7 strain of Escherichia colithrough MALDI-TOF/TOF MS in a top-down approach based onsix proteins (33). For one of the six, the YahO protein of unknownfunction, the exchange of a single amino acid was sufficient toenable a distinction of the pathogenic E. coli strains from the non-pathogenic strain.

SpeedVac 5 min HIFU

RapiGest Trypsin

15 minHIFU

Proteinextraction

Pelletresolubilization

Microbialculture

Trypsindigestion

nano-LC

Nano-LC separation,spotting on MALDI target

Identification ofbiomarker peptides

TOF/TOFmeasurement

ultrafleXtreme

FIG 2 Workflow for the preparation of tryptic digests of cell extracts from S.enterica subsp. arizonae, S. enterica subsp. enterica, and S. enterica subsp.houtenae and subsequent MALDI-TOF/TOF MS analysis. The cell extract usedfor classical biotyping is dried on a SpeedVac. After resolubilizing the pelletthrough the addition of RapiGest detergent under 5 min of high-intensityfocused ultrasound (HIFU) treatment, trypsin is added and the proteins aredigested for 15 min under the assistance of HIFU (in an ice water bath). Thepeptides are separated by nano-LC, mixed with HCCA matrix directly in thefraction collector, and spotted onto the MALDI target. MALDI-TOF/TOFanalysis is performed, and the resulting data are screened for biomarker pep-tide candidates using an Excel macro developed in house.

TABLE 1 Numbers of potential biomarker peptide masses found for S.enterica subsp. arizonae, S. enterica subsp. enterica, and S. enterica subsp.houtenae and numbers of peptides assigned to actual peptides inSalmonella proteins of the correct subspecies found by the MASCOTsearch using the NCBI databasea

Salmonella subspecies

No. of potentialbiomarkerpeptides

No. of assignedbiomarkerpeptides

S. enterica subsp. arizonae 51 29S. enterica subsp. enterica 48 22S. enterica subsp. houtenae 26 17a Data represent �200 ppm at an S/N ratio of 3.

Gekenidis et al.

4238 aem.asm.org Applied and Environmental Microbiology

on June 8, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

An alternative to the identification of proteins is the analysis ofnucleic acids by MALDI-TOF MS. In the so-called Sequenom ap-proach, similarly to the trypsin digestion step of the improvedproteomics workflow described in the present work, a base cleav-age step is integrated to generate more products for identificationthrough MALDI-TOF. For example, Syrmis et al. (34) detectedsingle nucleotide polymorphisms (SNPs) useful for differentia-tion of various methicillin-resistant Staphylococcus aureus strains.Such a targeted approach could be useful for the detection of se-rovar differences in, e.g., Salmonella strains for which no sequencedata are available. However, one should keep in mind that viableor dead contaminating bacteria might be detected by the requiredPCR amplification. Thus, the proteomics-based workflow offersan advantage by focusing on viable bacteria.

In conclusion, a fast and efficient proteomics workflow wasdeveloped in the present work which could be used for identifica-tion of subspecies-specific biomarker peptides of a multitude of

bacterial species. In future studies, unknown samples will bescreened for biomarker peptides by application of selected reac-tion monitoring (SRM) experiments to rapidly identify the sub-species to which they belong. The most suitable candidates wouldbe those biomarker peptides with a high peptide score, meaningthat the MS/MS peptide sequencing yielded satisfactory sequencecoverage. Such biomarker peptides displaying sequence coverageof about 70% or higher were found, e.g., in fructose-bisphosphatealdolase, cell division protein DamX, and DNA-binding tran-scriptional regulator PhoP for S. enterica subsp. arizonae, in a pu-tative outer membrane protein and a probable thiol peroxidase forS. enterica subsp. enterica, or in 60-kDa chaperonin, NAD(P)H:quinone oxidoreductase, and flagellar synthesis protein FlgN forS. enterica subsp. houtenae (see Table S2 in the supplemental ma-terial). Also, smaller peptides should be preferred since they areusually better suited from a chromatographic point of view inperforming LC-ESI-MS/MS. The SRM approach would allow a

TABLE 2 Selected examples for biomarker peptides which are discriminative for all three investigated subspeciesa

Salmonella subspecies Assigned proteinProteinmass (kDa)

Biomarkerpeptide mass (Da) Amino acid exchange(s) (subspecies)c

S. arizonae 6-Phosphogluconolactonase 36.4 3,004.52 � 0.03b A¡C, Y¡H, A¡V (S. enterica)A¡C, Y¡H, A¡V, E¡A (S. houtenae)

S. arizonae Arsenate reductase 13.4 1,420.81 � 0.00 E¡D (S. enterica*)E¡D, G¡S (S. houtenae*)

S. arizonae Ferric uptake regulator 17.0 3,291.42 � 0.01 Q¡H (S. enterica)E¡D, Q¡H (S. houtenae)

S. arizonae Membrane protein 15.7 3,051.48 � 0.02b L¡I, T¡A (S. enterica*)L¡I, M¡V, T¡A (S. houtenae*)

S. arizonae N-Ethylmaleimide reductase 39.5 2,590.17 � 0.02 K¡E, E¡A (S. enterica*)E¡A (S. houtenae*)

S. arizonae NADH dehydrogenase subunit G 100.0 3,009.44 � 0.01b H¡Q (S. enterica)A¡T, D¡G, H¡Q (S. houtenae)

S. enterica Cell division protein DamX 45.5 3,153.62 � 0.03b TT¡AP, TA¡KV (S. arizonae●)A¡T, T¡K, A¡T (S. houtenae)

S. enterica Fumarate hydratase 51.8 2,460.30 � 0.01b P¡S (S. arizonae)L¡M (S. houtenae*)

S. enterica H-NS histone family 15.5 1,979.88 � 0.01 MD¡IN (S. arizonae*)D¡G (S. houtenae*)

S. enterica Peptidyl-prolyl cis-trans-isomerase PpiD 58.2 2,187.04 � 0.00 V¡A, V¡T (S. arizonae*)V¡T (S. houtenae*)

S. enterica Probable thiol peroxidase 18.0 2,379.23 � 0.01 S¡N (S. arizonae*)T¡A (S. houtenae*)

S. enterica Putative outer membrane protein 12.8 3,224.54 � 0.02b L¡F, S¡R, S¡I, N¡Y (S. arizonae●)S¡T (S. houtenae*)

S. enterica RNase E 119.3 2,706.37 � 0.00b S¡T (S. arizonae*)P¡T, S¡T (S. houtenae*)

3,046.42 � 0.02b T¡A, S¡T (S. arizonae*)T¡A, H¡Y (S. houtenae*)

S. houtenae Flagellar synthesis protein FlgN 16.0 3,102.61 � 0.05 S¡A (S. arizonae)N¡D (S. enterica*)

S. houtenae Isoprenoid biosynthesis protein 22.9 2,763.46 � 0.02 S¡A (S. arizonae*)E¡D (S. enterica*)

S. houtenae NlpB/DapX lipoprotein 36.9 2,504.25 � 0.01 D¡E, V¡P (S. arizonae)V¡A (S. enterica*)

S. houtenae Outer membrane protein A 38.4 3,521.70 � 0.02b A¡T, Y¡GA, T¡Y (S. arizonae*)A¡V, Y¡GP (S. enterica*)

a The measured peptide masses are given as the arithmetic mean calculated from the values of the three replicates of the subspecies, and the corresponding standard deviations arepresented. Specific amino acids in the biomarker peptides of a subspecies and their exchanges in peptides of the other subspecies are given. All numbers were rounded to the seconddigit after the comma. The NCBI database was used.b Biomarker peptide found in all investigated ppm windows (�25, �50, and �100 ppm).c *, a sequence analogous to the biomarker peptide was observed; ●, a shorter or longer peptide sequence containing at least one relevant amino acid exchange was observed.

Tryptic Peptides for Discrimination of Subspecies

July 2014 Volume 80 Number 14 aem.asm.org 4239

on June 8, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

rapid and cost-effective screening and classification of unknownbacterial samples. Finally, prior studies (30) have documentedSalmonella serovar-level differences in protein profiles and havetraced the responsible amino acid exchanges. Thus, the procedurepresented here could be further applied to identify biomarkerpeptides discriminatory of different serovars within a given sub-species.

ACKNOWLEDGMENTS

We are indebted to the protein analysis group of the Functional GenomicsCenter Zurich for valuable discussions, Brion Duffy and KerstinBrankatschk for supplying Salmonella enterica subsp. enterica, and DanielBaumgartner for his valuable help in developing the Excel macro.

REFERENCES1. Claydon MA, Davey SN, Edwards-Jones V, Gordon DB. 1996. The rapid

identification of intact microorganisms using mass spectrometry. Nat.Biotechnol. 14:1584 –1586. http://dx.doi.org/10.1038/nbt1196-1584.

2. Dieckmann R, Strauch E, Alter T. 2010. Rapid identification and char-acterization of Vibrio species using whole-cell MALDI-TOF mass spec-trometry. J. Appl. Microbiol. 109:199 –211. http://dx.doi.org/10.1111/j.1365-2672.2009.04647.x.

3. Fenselau C, Demirev PA. 2001. Characterization of intact microorgan-isms by MALDI mass spectrometry. Mass Spectrom. Rev. 20:157–171.http://dx.doi.org/10.1002/mas.10004.

4. Holland RD, Wilkes JG, Rafii F, Sutherland JB, Persons CC, Voorhees KJ, LayJO, Jr. 1996. Rapid identification of intact whole bacteria based on spectral pat-terns using matrix-assisted laser desorption/ionization with time-of-flight massspectrometry. Rapid Commun. Mass Spectrom. 10:1227–1232. http://dx.doi.org/10.1002/(SICI)1097-0231(19960731)10:101227::AID-RCM6593.0.CO;2-6.

5. Krishnamurthy T, Ross PL, Rajamani U. 1996. Detection of patho-genic and non-pathogenic bacteria by matrix-assisted laser desorp-tion/ionization time-of-flight mass spectrometry. Rapid Commun.Mass Spectrom. 10:883– 888. http://dx.doi.org/10.1002/(SICI)1097-0231(19960610)10:8883::AID-RCM5943.0.CO;2-V.

6. Valentine N, Wunschel S, Wunschel D, Petersen C, Wahl K. 2005.Effect of culture conditions on microorganism identification by matrix-assisted laser desorption ionization mass spectrometry. Appl. Environ.Microbiol. 71:58 – 64. http://dx.doi.org/10.1128/AEM.71.1.58-64.2005.

7. Wunschel DS, Hill EA, McLean JS, Jarman K, Gorby YA, Valentine N,Wahl K. 2005. Effects of varied pH, growth rate and temperature usingcontrolled fermentation and batch culture on matrix assisted laser desorp-tion/ionization whole cell protein fingerprints. J. Microbiol. Methods 62:259 –271. http://dx.doi.org/10.1016/j.mimet.2005.04.033.

8. Wunschel SC, Jarman KH, Petersen CE, Valentine NB, Wahl KL,Schauki D, Jackman J, Nelson CP, White EV. 2005. Bacterial analysis byMALDI-TOF mass spectrometry: an inter-laboratory comparison. J. Am.Soc. Mass Spectrom. 16:456 – 462. http://dx.doi.org/10.1016/j.jasms.2004.12.004.

9. Kliem M, Sauer S. 2012. The essence on mass spectrometry based micro-bial diagnostics. Curr. Opin. Microbiol. 15:397– 402. http://dx.doi.org/10.1016/j.mib.2012.02.006.

10. van Veen SQ, Claas ECJ, Kuijper EJ. 2010. High-throughput identifica-tion of bacteria and yeast by matrix-assisted laser desorption ionization-time of flight mass spectrometry in conventional medical microbiologylaboratories. J. Clin. Microbiol. 48:900 –907. http://dx.doi.org/10.1128/JCM.02071-09.

11. Seng P, Drancourt M, Gouriet F, La Scola B, Fournier PE, Rolain JM,Raoult D. 2009. Ongoing revolution in bacteriology: routine identifica-tion of bacteria by matrix-assisted laser desorption ionization time-of-flight mass spectrometry. Clin. Infect. Dis. 49:543–551. http://dx.doi.org/10.1086/600885.

12. Sogawa K, Watanabe M, Sato K, Segawa S, Ishii C, Miyabe A, MurataS, Saito T, Nomura F. 2011. Use of the MALDI BioTyper system withMALDI-TOF mass spectrometry for rapid identification of microorgan-isms. Anal. Bioanal. Chem. 400:1905–1911. http://dx.doi.org/10.1007/s00216-011-4877-7.

13. Cherkaoui A, Hibbs J, Emonet S, Tangomo M, Girard M, Francois P,Schrenzel J. 2010. Comparison of two matrix-assisted laser desorptionionization-time of flight mass spectrometry methods with conventional

phenotypic identification for routine identification of bacteria to the spe-cies level. J. Clin. Microbiol. 48:1169 –1175. http://dx.doi.org/10.1128/JCM.01881-09.

14. Bizzini A, Durussel C, Bille J, Greub G, Prod’hom G. 2010. Performanceof matrix-assisted laser desorption ionization-time of flight mass spec-trometry for identification of bacterial strains routinely isolated in a clin-ical microbiology laboratory. J. Clin. Microbiol. 48:1549 –1554. http://dx.doi.org/10.1128/JCM.01794-09.

15. Croxatto A, Prod’hom G, Greub G. 2012. Applications of MALDI-TOFmass spectrometry in clinical diagnostic microbiology. FEMS Microbiol.Rev. 36:380 – 407. http://dx.doi.org/10.1111/j.1574-6976.2011.00298.x.

16. Sandrin TR, Goldstein JE, Schumaker S. 2013. MALDI TOF MS profilingof bacteria at the strain level: a review. Mass Spectrom. Rev. 32:188 –217.http://dx.doi.org/10.1002/mas.21359.

17. López-Ferrer D, Capelo JL, Vázquez J. 2005. Ultra fast trypsin digestionof proteins by high intensity focused ultrasound. J. Proteome Res. 4:1569 –1574. http://dx.doi.org/10.1021/pr050112v.

18. Sauer S, Freiwald A, Maier T, Kube M, Reinhardt R, Kostrzewa M,Geider K. 2008. Classification and identification of bacteria by mass spec-trometry and computational analysis. PLoS One 3:e2843. http://dx.doi.org/10.1371/journal.pone.0002843.

19. Barbuddhe SB, Maier T, Schwarz G, Kostrzewa M, Hof H, Domann E,Chakraborty T, Hain T. 2008. Rapid identification and typing of Listeriaspecies by matrix-assisted laser desorption ionization-time of flight massspectrometry. Appl. Environ. Microbiol. 74:5402–5407. http://dx.doi.org/10.1128/AEM.02689-07.

20. Dubois D, Leyssene D, Chacornac JP, Kostrzewa M, Schmit PO, TalonR, Bonnet R, Delmas J. 2010. Identification of a variety of Staphylococcusspecies by matrix-assisted laser desorption ionization-time of flight massspectrometry. J. Clin. Microbiol. 48:941–945. http://dx.doi.org/10.1128/JCM.00413-09.

21. Mellmann A, Cloud J, Maier T, Keckevoet U, Ramminger I, Iwen P,Dunn J, Hall G, Wilson D, LaSala P, Kostrzewa M, Harmsen D. 2008.Evaluation of matrix-assisted laser desorption ionization-time-of-flightmass spectrometry in comparison to 16S rRNA gene sequencing for spe-cies identification of nonfermenting bacteria. J. Clin. Microbiol. 46:1946 –1954. http://dx.doi.org/10.1128/JCM.00157-08.

22. Wensing A, Zimmermann S, Geider K. 2010. Identification of the cornpathogen Pantoea stewartii by mass spectrometry of whole-cell extractsand its detection with novel PCR primers. Appl. Environ. Microbiol. 76:6248 – 6256. http://dx.doi.org/10.1128/AEM.01032-10.

23. Ghyselinck J, Van Hoorde K, Hoste B, Heylen K, De Vos P. 2011. Evalu-ation of MALDI-TOF MS as a tool for high-throughput dereplication. J. Mi-crobiol. Methods 86:327–336. http://dx.doi.org/10.1016/j.mimet.2011.06.004.

24. Pavlovic M, Huber I, Konrad R, Busch U. 2013. Application of MALDI-TOF MS for the identification of food borne bacteria. Open Microbiol. J.7:135–141. http://dx.doi.org/10.2174/1874285801307010135.

25. Reich M, Bosshard PP, Stark M, Beyser K, Borgmann S. 2013. Speciesidentification of bacteria and fungi from solid and liquid culture media byMALDI-TOF mass spectrometry. J. Bacteriol. Parasitol. http://dx.doi.org/10.4172/2155-9597.S5-002.

26. Sparbier K, Weller U, Boogen C, Kostrzewa M. 2012. Rapid detection ofSalmonella sp. by means of a combination of selective enrichment brothand MALDI-TOF MS. Eur. J. Clin. Microbiol. Infect. Dis. 31:767–773.http://dx.doi.org/10.1007/s10096-011-1373-0.

27. Brankatschk K, Blom J, Goesmann A, Smits THM, Duffy B. 2011.Genome of a European fresh-vegetable food safety outbreak strain of Sal-monella enterica subsp. enterica serovar Weltevreden. J. Bacteriol. 193:2066. http://dx.doi.org/10.1128/JB.00123-11.

28. Madonna AJ, Basile F, Ferrer I, Meetani MA, Rees JC, Voorhees KJ. 2000.On-probe sample pretreatment for the detection of proteins above 15KDa from whole cell bacteria by matrix-assisted laser desorption/ion-ization time-of-flight mass spectrometry. Rapid Commun. Mass Spectrom.14:2220–2229. http://dx.doi.org/10.1002/1097-0231(20001215)14:232220::AID-RCM1553.0.CO;2-4.

29. Dieckmann R, Helmuth R, Erhard M, Malorny B. 2008. Rapid classifi-cation and identification of salmonellae at the species and subspecies levelsby whole-cell matrix-assisted laser desorption ionization-time of flightmass spectrometry. Appl. Environ. Microbiol. 74:7767–7778. http://dx.doi.org/10.1128/AEM.01402-08.

30. Dieckmann R, Malorny B. 2011. Rapid screening of epidemiologicallyimportant Salmonella enterica subsp. enterica serovars by whole-cell ma-

Gekenidis et al.

4240 aem.asm.org Applied and Environmental Microbiology

on June 8, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

trix-assisted laser desorption ionization-time of flight mass spectrometry.Appl. Environ. Microbiol. 77:4136 – 4146. http://dx.doi.org/10.1128/AEM.02418-10.

31. Fagerquist CK, Miller WG, Harden LA, Bates AH, Vensel WH, WangG, Mandrell RE. 2005. Genomic and proteomic identification of a DNA-binding protein used in the “fingerprinting” of Campylobacter species andstrains by MALDI-TOF-MS protein biomarker analysis. Anal. Chem. 77:4897– 4907. http://dx.doi.org/10.1021/ac040193z.

32. Fagerquist CK, Bates AH, Heath S, King BC, Garbus BR, Harden LA,Miller WG. 2006. Sub-speciating Campylobacter jejuni by proteomic anal-ysis of its protein biomarkers and their post-translational modifications. J.Proteome Res. 5:2527–2538. http://dx.doi.org/10.1021/pr050485w.

33. Fagerquist CK, Garbus BR, Miller WG, Williams KE, Yee E, Bates AH,Boyle S, Harden LA, Cooley MB, Mandrell RE. 2010. Rapid identifica-tion of protein biomarkers of Escherichia coli O157:H7 by matrix-assistedlaser desorption ionization-time-of-flight-time-of-flight mass spectrom-etry and top-down proteomics. Anal. Chem. 82:2717–2725. http://dx.doi.org/10.1021/ac902455d.

34. Syrmis MW, Moser RJ, Whiley DM, Vaska V, Coombs GW, NissenMD, Sloots TP, Nimmo GR. 2011. Comparison of a multiplexedMassARRAY system with real-time allele-specific PCR technology forgenotyping of methicillin-resistant Staphylococcus aureus. Clin. Micro-biol. Infect. 17:1804 –1810. http://dx.doi.org/10.1111/j.1469-0691.2011.03521.x.

Tryptic Peptides for Discrimination of Subspecies

July 2014 Volume 80 Number 14 aem.asm.org 4241

on June 8, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from